Agricultural productivity is often compromised by plant diseases, particularly in staple crops such as maize. There is an increased usage of deep learning models, such as convolutional neural networks (CNNs), as an effective way for image-based disease diagnosis. This paper presents a hybrid approach that combines a Deep CNN and a custom CNN to improve accuracy in maize disease detection. Here, MobileNet, a lightweight pre-trained Deep CNN, is considered. The MobileNet component leverages pre-trained weights from ImageNet to extract deep semantic features. A computationally efficient custom CNN is designed that complements MobileNet by learning task-specific representations. For evaluating the performance of maize crops in the experiments, images are obtained from the Plant Village dataset. The dataset comprised images representing various conditions, including healthy and diseased leaves. Experimental results on the maize leaf disease dataset demonstrate significant improvements in classification accuracy, precision, and recall compared to conventional single-model approaches.